Podcast Clipper ( Subscut )

Use this skill when the user wants to turn a long podcast, interview, webinar, or talking-head video into multiple short clips for TikTok, Reels, or YouTube...

Security Scan

The skill's code, runtime instructions, and requested environment variables are coherent with a podcast-to-clips wrapper for Subscut's API and do not request unrelated credentials or perform unexpected actions.

Purpose & Capability

Name/description (podcast-to-clips) match the files and runtime behavior: a small Node CLI that POSTs the video URL and rendering options to a Subscut /podcast-to-clips endpoint. Requested binaries (npm) and env var (SUBSCUT_API_KEY) are appropriate for this purpose. Minor metadata inconsistencies exist (registry lists SUBSCUT_API_BASE_URL as required while SKILL.md treats it optional; package/version strings differ), but these are documentation issues rather than functional mismatches.

Instruction Scope

SKILL.md and the bundled script limit actions to validating inputs and calling the declared API endpoint. The instructions reference only the declared env vars and the CLI wrapper; they do not read other system files, secrets, or arbitrary paths, nor do they send data to unexpected hosts.

Install Mechanism

There is no download/install-from-URL step; the skill ships a small JS script and examples. No extracted archives or external installers are used. The script relies on Node/npm being available, which aligns with the declared required binary.

Credentials

The skill requires a single service credential (SUBSCUT_API_KEY) and an optional base URL. This is proportionate to a third-party API wrapper. Note: registry metadata listed SUBSCUT_API_BASE_URL as required while SKILL.md marks it optional (defaulting to https://subscut.com) — an innocuous documentation mismatch but worth noticing. No other secrets or unrelated credentials are requested.

Persistence & Privilege

The skill is not always-enabled and does not request elevated platform privileges. It does not modify other skills' config or persist new agent-wide credentials. Autonomous invocation remains enabled (platform default) but is not combined with other red flags.

Assessment

This package appears to do exactly what it says: a small Node CLI that calls Subscut's /podcast-to-clips endpoint using SUBSCUT_API_KEY. Before installing, verify the SUBSCUT_API_KEY comes from the legitimate Subscut service and grant it only the minimum permissions required. Be aware of minor doc mismatches (base URL marked required in registry metadata but optional in SKILL.md, and package/version strings differ); these are likely bookkeeping issues but you may want to confirm the published skill matches the official Subscut marketplace entry. If you prefer to be extra cautious, run the script locally with a test API key or inspect network traffic (or set SUBSCUT_API_BASE_URL to a local proxy) to confirm it only contacts the expected host.

Generate Podcast Clips

Use this skill to convert a long-form spoken video into multiple short clips through the Subscut /podcast-to-clips API.

What This Skill Does

The skill is an opinionated wrapper around the API outcome:

• extracts up to 20 strong short clips from a long-form video
• favors viral and high-retention spoken moments
• adds captions with selectable styles
• supports two render formats: dynamic (auto-reframing) and hook_frame (original frame + title card)
• returns titles, scores, and rendered clip URLs

Think in outcomes, not transport:

• Good framing: "Extract viral short-form content from this podcast"
• Bad framing: "Call some video API"

When To Use

Use this skill when:

• the input is a long podcast, interview, webinar, or talking-head video
• the user wants growth, repurposing, shorts, reels, or TikTok content
• the user wants minimal manual editing

Avoid this skill when:

• the source is already short-form
• the content is mostly non-speech
• the user wants manual, frame-by-frame editing decisions

Do not use it as a generic video editing tool.

Input Contract

Use this compact input shape when planning or explaining the tool call:

{
  "video_url": "https://example.com/video.mp4",
  "max_clips": 5,
  "style": "viral",
  "format": "dynamic",
  "captions": true,
  "clip_duration": {
    "min": 20,
    "max": 60
  }
}

Field Reference

Field	Type	Default	Notes
video_url	string	—	Required. Any HTTP/HTTPS URL. YouTube, direct MP4, Google Drive.
max_clips	integer	5	Range: 1–20. Short videos (≤3 min) are capped at 2 clips automatically.
style	string	"viral"	Caption style. See styles below.
format	string	"dynamic"	Render format. See formats below.
captions	boolean	true	Whether to burn in captions.
clip_duration.min	integer	20	Minimum clip length in seconds. Floor: 10s.
clip_duration.max	integer	60	Maximum clip length in seconds. Ceiling: 60s. Must be ≥ min.

Caption Styles (style)

Value	Description
viral	Bold animated-word captions (MrBeast style). Default.
beast	Alias for viral.
hormozi	Alias for leon. Single highlighted word, clean font.
leon	Single highlighted word, clean font.
clean	Plain white subtitles, no animation.
minimal	Alias for clean.

Render Formats (format)

Value	Description
dynamic	Auto-detects split-screen vs. solo framing, reframes to 9:16. Default.
hook_frame	Preserves original video frame, adds a title card at the top, captions at the bottom.

Use hook_frame when the video is already vertical or the user wants the title displayed prominently. Use dynamic (default) for horizontal/landscape podcasts with one or two speakers.

Output Contract

Expect JSON in this shape:

{
  "clips": [
    {
      "video_url": "https://...",
      "title": "Why Most Founders Get This Wrong",
      "score": 0.92,
      "start": 142.5,
      "end": 198.3
    }
  ]
}

score is a 0–1 float representing clip virality confidence. Higher is better.

CLI Entry Point

Once the skill is installed, the agent runtime should invoke the bundled CLI wrapper:

npm --prefix .agents/skills/generate-podcast-clips run generate-podcast-clips -- \
  --video-url "https://example.com/podcast.mp4" \
  --max-clips 5 \
  --clip-style viral \
  --format dynamic \
  --captions true \
  --min-clip-duration 20 \
  --max-clip-duration 60

Required environment variables:

• SUBSCUT_API_KEY or --api-key

Default base URL:

• https://subscut.com

Install Model

This skill is meant to be published once by Subscut and then installed by users through the marketplace.

End-user flow:

1. Install the published skill from ClawHub / OpenClaw.
2. Set SUBSCUT_API_KEY.
3. Let the agent call the skill when it needs to turn a long-form spoken video into short clips.

The skill should not ask users to publish or package anything themselves.

Agent Workflow

1. Confirm the source is a long-form spoken video.
2. Prefer the CLI wrapper over hand-written curl.
3. Keep the request simple unless the user asks for custom clip counts, durations, style, or format.
4. Default format to dynamic unless the user explicitly wants a title card (hook_frame).
5. Return the resulting clip URLs, titles, scores, and timestamps.
6. If the API fails, surface the status and response body clearly.
7. Keep a human in the loop before final publishing if the downstream workflow is public.

Natural-Language Triggers

Likely user intents that should map to this skill:

• "Turn this podcast into reels"
• "Make shorts from this interview"
• "Repurpose my long video into clips"
• "Find the viral moments in this podcast"
• "Generate YouTube Shorts from this episode"
• "Make clips with a title card at the top"
• "Give me clean subtitle clips"
• "Extract 10 clips from this webinar"

Notes

• The API route is POST /podcast-to-clips.
• style maps to the API style field. clip-style is used as the CLI flag name.
• format controls render layout: dynamic (default) or hook_frame.
• clip_duration.min and clip_duration.max let callers control clip length window.
• Short videos (≤3 min) are automatically capped at 2 clips regardless of max_clips.
• The script is executed through the skill-local npm script.
• The skill is intentionally opinionated and keeps the parameter surface small for better agent usage.
• This package is for installation and runtime usage, not for publisher-only deployment steps.